Skip to content
Model infoModelsSnapshot: 15 Oct 2026

xAI / Grok Voice

Grok Voice Think Fast 2.0

Bidirectional voice and text conversations with configurable reasoning and tool calls.

availableClosed weightsSpeech generationAgentsReasoning
Official documentation (opens in a new tab)Reviewed 12 Oct 2026
15 of 23 specification fields collected
Context windowNot collectedAwaiting source review
Maximum outputNot collectedAwaiting source review
Input billingSee ratesxAI API
Output billingSee ratesPublished units and conditions below

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
Not collected
Maximum output
Not collected
Structured output
Not collected
Streaming
SupportedSource for Streaming (opens in a new tab)

Bidirectional WebSocket audio and text.

Batch processing
Not collected
Fine-tuning
Not collected
Knowledge cutoff
Not collected
Release date
2026-07Source for Release date (opens in a new tab)

Release month from the provider changelog; exact day not collected.

Parameters
Not collected
Architecture
Not collected
Languages
20+ supported languages; automatic input-language detection.Source for Languages (opens in a new tab)
Supported tools
file_search, web_search, x_search, mcp, functionSource for Supported tools (opens in a new tab)
Audio formats
audio/pcm, audio/pcmu, audio/pcma, audio/opusSource for Audio formats (opens in a new tab)
Audio sample rates
8,000 / 16,000 / 22,050 / 24,000 / 32,000 / 44,100 / 48,000 Hz (PCM)Source for Audio sample rates (opens in a new tab)
Supported API routes
wss://api.x.ai/v1/realtimeSource for Supported API routes (opens in a new tab)
Context caching
SupportedSource for Context caching (opens in a new tab)

Documented on at least one recorded endpoint. See Caching for host, lifecycle and billing restrictions.

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

API model ID
grok-voice-think-fast-2.0
Access
API account required. Model access and quotas depend on the team and location.
Regions
Global endpoint; processing location is not guaranteed.

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 12 Oct 2026
CategoryUsageRateUnit
Speech to speechConversation audio1$0.08minute
Speech to speechText input2$0.004text input

Rate conditions

  1. Speech-to-speech rate; the provider also expresses this as $4.80/hour, which is the same rate and not an additional charge.
  2. Separate charge for a text input in a voice conversation.
Pricing source (opens in a new tab)

Data & enterprise terms

Provider documents no training on API inputs or outputs without explicit permission, and default 30-day audit retention. Team-level zero data retention changes available features; consult the linked policy for scope and exceptions.

Read the service policy (opens in a new tab)

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

xAI API

grok-voice-think-fast-2.0
Documented
Support evidence (opens in a new tab)
How it works
Automatic prefix cachingSource for How it works (opens in a new tab)
Minimum reusable input
Not collected
Lifetime & refresh
No guaranteed duration. Entries can be evicted because of load or server restarts.Source for Lifetime & refresh (opens in a new tab)
How to enable it
Keep the initial messages unchanged. Use x-grok-conv-id, or prompt_cache_key in Responses, to improve routing consistency.Source for How to enable it (opens in a new tab)
Usage fields
Chat: usage.prompt_tokens_details.cached_tokens. Responses: usage.input_tokens_details.cached_tokens.Source for Usage fields (opens in a new tab)
Billing rules
Cached input uses the rate for the selected model and endpoint. Uncached input and output are billed separately; observe any prompt-length tiers.Source for Billing rules (opens in a new tab)
Limits & data boundaries
Hits are best effort. Editing or reordering early messages breaks prefix reuse. Streaming and tool-call history are supported.Source for Limits & data boundaries (opens in a new tab)

No verified cache rate for this endpoint. An empty rate is not a zero charge.

Caching reviewed 12 Oct 2026

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Developer-documented proprietary hosted Grok release. No public weights identified for this release in the reviewed documentation; public Grok 1 and Grok 2 base-weight releases are excluded.

Classification evidence (opens in a new tab)

Lifecycle dates

grok-voice-latest alias scheduled to route to this release5 Aug 2026Source (opens in a new tab)

Sources reviewed